Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
80
Asymmetric Adjustment of Rice Export Prices: The Case
of Thailand and Vietnam
Atanu Ghoshray
University of Bath, UK
Abstract This paper attempts to model the price relationship between two leading exporters
of rice, being Thailand and Vietnam. The motivation of such research is to reveal whether
prices are integrated and determine whether any causal relationships exist among prices for
rice. Given the perception that exists regarding the imperfectly competitive nature of the rice
market, especially in the high quality rice segment, the paper aims to test for the presence of
cointegration in the presence of asymmetric adjustment. The results conclude that the nature
of asymmetry is captured by the M-TAR model which suggests that the path of adjustment to
the long run equilibrium relation is relatively faster when the price differential is decreasing
than opposed the case when it is increasing.
Keywords: Cointegration, TAR, M-TAR, consistent threshold, rice, asymmetric price
Adjustment
JEL Classification: Q13, Q17, C22
1. Introduction
Issues such as price transmission and market integration have been of significant interest to
economists which has lead to a large volume of literature on the study of international price
relationships in agricultural commodity markets. If a commodity market were to be
integrated, the prices charged by major exporters of that commodity will be expected to move
together over time through arbitrage or substitution or both. In other words, the export prices
of the commodity will not diverge too far from each other if a price change by one major
exporter is to be followed by a gradual similar price change by the other major exporter. In
this way, by examining the relationships between prices over time, the relevant market can be
defined. From an econometric point of view, this would imply that prices of the commodity
should be cointegrated.
Rice is an internationally traded commodity and is the staple food for nearly half of the
world’s population (Chen et al., 2006). The world rice market is thin given the small amount
of rice traded relative to production (Nielsen 2003). During the past decade approximately
85% of the rice exports have emanated from six countries, being Thailand, USA, Vietnam,
India, Pakistan and China1. The rice market has the highest level of policy distortion in
agricultural commodity markets (Rakotoarisoa 2006). The interventions made by the
government with an aim towards achieving self sufficiency, is a contributing factor towards
rice being a residual market (Siamwalla and Haykin 1983). The role of governments can be
important as they can develop a strategic response to the rice policies of other countries. One
possible strategy might involve matching a country’s price decreases but not price increases.
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
81
Further, the demand for rice is affected by the specific types and qualities of rice (Cramer et
al., 1993). Substitution among the various rice qualities is limited by the taste preferences of
various importing countries. For instance high quality rice (which would imply a lower
percentage of broken kernels) commands a premium and is strongly preferred in developed
countries and in the higher income groups of developing countries (Cramer et al., 1993). In
particular, the presence of different qualities of an agricultural commodity can be a factor for
asymmetric price adjustment to occur (Ghoshray 2002). Traditionally Vietnam has been
exporting rice which is of intermediate and low quality which sells at a significant discount to
its Thai counterparts (Nielsen 2003). However, in recent years, Vietnam has been exporting
high quality rice and has been competing with high quality rice exports of other major
exporters.
Past studies have tried to characterize the nature of the rice market. Karp and Perloff (1989)
conclude that the rice market is perfectly competitive. However, in a later study, Yumkella et
al., (1994) find that only the high quality rice market is non-competitive. Goletti and Babu
(1994) observe that if the market were to be characterized as perfectly competitive, then price
increases should be transmitted to the same extent as price decreases. However, in the case of
the rice export market, the price transmission is likely to be asymmetric, given that the
market has a structure more consistent with an imperfectly competitive market. If the
underlying process of adjustment is asymmetric, the standard test for cointegration and its
extensions are mis-specified (Enders and Siklos 2001). Recent developments in time series
analysis have recognized the potential for threshold type adjustments in error correction
models. This issue has been frequently overlooked in the past literature on agricultural
commodity markets and has not been applied in the case of the rice market and therefore
deserves further attention.
Thailand is the largest exporter of rice and although a number of private companies manage
the exports of rice from Thailand, the government still exercises some control over rice
exports because of the important role that rice plays in the Thai economy (Rakotoarisoa
2006). In the late 1980s Vietnam adopted reforms in agriculture which triggered an expansion
in rice production and helped to transform Vietnam from a rice importer to a rice exporter
(Minot and Goletti 1998). In contrast to Thailand, all rice exports from Vietnam are managed
by the state owned Vietnam Food Corporation (Rakotoarisoa 2006). Over the period analyzed
in this study, Thailand has been the dominant exporter accounting for 32% of total world
exports followed by Vietnam with 15% in second place.2
The objective of this paper is to analyze the dynamics between the rice export prices of the
Thailand and Vietnam. The motivation of such research is to reveal whether prices are
integrated and determine whether any causal relationships exist among prices for rice. Given
the perception that exists regarding the imperfectly competitive nature of the rice market,
especially in the high quality segment, the paper aims to test for the presence of cointegration
in the presence of asymmetric error correction across different qualities of rice. The different
qualities of rice are divided into three groups, being ‘high’, ‘medium’ and ‘low’ quality rice.
The correspondence between error correction models representing cointegrating relationships
and autoregressive models of an error term allow us to apply the method put forward by
Enders and Siklos (2001), that is, the threshold autoregressive (TAR) and momentumthreshold autoregressive (M-TAR) method of adjustment. This type of asymmetric price
behavior differs from the scenario where both prices move together and any transitory
deviation in ‘levels’ or ‘rates of change’ is corrected in a symmetric manner.
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
82
This paper is organized as follows: Section 2 describes the econometric model followed by a
description of the data and the empirical results in Section 3. Section 4 draws the conclusions
from the analysis.
2. Econometric Model
At the onset, the Engle and Granger (1987) two-step method is employed to test for
cointegration between the two export prices to be considered in this study, say prices P1t and
P2 t . This entails using ordinary least squares (OLS) to estimate the long-run relation of the
two prices given by the equation below:
P1t = α + βP2 t + ε t
(1)
where P1t and P2 t are non-stationary I(1) prices, α and β are the estimated parameters. The
arbitrary constant α accounts for the differential (transfer costs and quality differences), β
denotes the price transmission elasticity and ε t is the error term which may be serially
correlated. The second step advocates a Dickey-Fuller test on the estimated residuals εˆt of
(1) as follows:
Δεˆt = γεˆt −1 + ωt
(2)
where ωt is a white noise error term.3 Rejecting the null hypothesis (H 0 : γ = 0) of no
cointegration implies that the residuals of (2) are stationary. Thus (1) is like an attractor such
that its pull is strictly proportional to the absolute value of εˆt .
However, Enders and Siklos (2001) argue that the test for cointegration and its extensions are
mis-specified if adjustment is asymmetric. They consider an alternative specification, called
the threshold autoregressive (TAR) model, such that (2) can be written as:
Δεˆt = I t γ 1εˆt −1 + (1 − I t )γ 2εˆt −1 + ωt
(3)
where I t is the Heaviside indicator function such that:
⎧1 if εˆt-1 ≥ τ
It = ⎨
⎩0 if εˆt-1 < τ
(4)
This specification allows for asymmetric adjustment. If the system is convergent, then the
long run equilibrium value of the sequence is given by εˆt = τ . The sufficient conditions for
the stationarity of εˆt are γ 1 < 0 , γ 2 < 0 and (1 + γ 1 )(1 + γ 2 ) < 1 (Petrucelli and Woolford
1984). In this case if εˆt −1 is above its long run equilibrium value, then adjustment is at the
rate γ 1 and if εˆt −1 is below long run equilibrium then adjustment is at the rate γ 2 . The
adjustment would be symmetric if γ 1 = γ 2 . However, if the null hypothesis H 0 : (γ 1 = γ 2 ) is
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
83
rejected then using the TAR model we can capture signs of asymmetry. If for example,
− 1 < γ 1 < γ 2 < 0 , then the negative phase of the εˆt series will tend to be more persistent than
the positive phase. In the above case it is necessary to estimate the value of the threshold that
will be equal to the cointegrating vector. A method of searching for a consistent estimate of
the threshold was undertaken by using a method proposed by Chan (1993).4
Enders and Siklos (2001) suggest a further alternative such that the threshold depends on the
previous periods change in εˆt instead on the level of εˆt . The Heaviside Indicator in this case
can be set to the Momentum-Heaviside Indicator as follows:
⎧1 if Δεˆt-1 ≥ τ
It = ⎨
⎩0 if Δεˆt-1 < τ
(5)
In this case the series εˆt exhibits more momentum in one direction than the other. The model
given by (3) along with equation (5) depicts the momentum threshold autoregression (MTAR) model. The M-TAR model can be used to capture a different type of asymmetry. If for
example, | γ 1 | < | γ 2 | , the M-TAR model exhibits little adjustment for positive Δεˆt −1 but
substantial decay for negative Δεˆt −1 . Alternatively, increases tend to persist, but decreases
tend to revert quickly back to the attractor irrespective of where disequilibrium is relative to
the attractor. The threshold is estimated using Chan’s methodology as before.4
To implement in this test the case of the TAR or M-TAR adjustment the Heaviside Indicator
function is set according to equation (4) or equation (5) respectively and estimate equation (3)
accordingly. The Φ -statistic for the null hypothesis of stationarity of εˆt , i.e.
H 0 : (γ 1 = γ 2 = 0) is recorded. The value of the Φ -statistic is compared to the critical values
computed by Enders and Granger (1998). If we can reject the null hypothesis, it is possible to
test for asymmetric adjustment since γ 1 and γ 2 converge to multivariate normal distributions
(Tong 1990). The F statistic is used to test for the null hypothesis of symmetric adjustment,
that is, H 0 : (γ 1 = γ 2 ) . Diagnostic checking of the residuals are undertaken to ascertain
whether the ωt series is a white noise process using the Ljung-Box Q tests. If the residuals
are correlated, equation (3) needs to be re-estimated in the form:
p
Δεˆt = I t γ 1εˆt −1 + (1 − I t )γ 2εˆt −1 + ∑ φi Δεˆt −1 + ωt
(6)
i =1
Equation (6) is comparable to (3) except that it incorporates lagged first differences of the
dependent variable to correct for the autocorrelation in the error term ω t .
The positive finding of cointegration with threshold adjustment would justify the estimation
of the following error correction model with threshold adjustment. The model takes the form:
p
p
i =1
i =1
ΔP1t = μ1 ECM t+−1 + μ2 ECM t−−1 + ∑ψ i ΔP1t −i + ∑ δ i ΔP2t −i + u1t
(7)
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
ΔP2t = π 1 ECM
+
t −1
+ π 2 ECM
−
t −1
p
p
i =1
i =1
+ ∑ηi ΔP1t −i + ∑ λi ΔP2t −i + u2t
84
(8)
where P1 and P2 denote the export prices of the two countries chosen in this study, ECM
refers to the error correction term and both u1t and u2t are white noise errors. The number of
lags p is determined using the SBC.
3. Data and Empirical Analysis
The data used for this analysis are monthly average export price quotations (FOB) from
August 1997 to November 2006. The rice prices used in this study include those of Thailand
and Vietnam. These include 5% Broken, 15% Broken and 35% Broken Thai rice; and
Vietnam 5% Broken, 15% Broken and 25% Broken rice. The 5% broken rice is included to
indicate high quality, 15% to indicate intermediate and the 25–35% broken to depict low
quality. The Thai price data was obtained from the Thailand Grain and Feed Weekly Rice
Price Update from the U.S. Embassy in Bangkok. The Vietnamese prices were obtained from
the Creed Rice Market Report, Creed Rice Co. Inc., Houston, Texas, USA. All prices are
quoted in U.S. dollars per ton. The subsequent analysis of the data is carried out on the
logarithm of prices. Figure 1, Panels A, B and C illustrate the rice export prices from
Thailand and Vietnam for the three different qualities.
The prices were initially tested for their order of integration using the ADF test, Elliot
Rothenberg and Stock (hereafter ERS) test (Elliot et al., 1996) and the Ng and Perron (1995)
test. Table 1 below presents the results of the unit root tests for each of the price series. The
ADF unit root test results for the variables in levels and growth form conclude that the prices
are non-stationary I(1). These results are supported by the more powerful ERS and Ng-Perron
tests.
In the analysis that follows a test is made for cointegration between Thai and Vietnam prices.
In each case different types of rice, differentiated by processing and quality of milled grain
are studied. Following the Engle Granger (1987) methodology, the first step entails
estimating the long run equilibrium relationship given by equation (1) and then conducting
the ADF test on the residuals of (1). The results of the test between different types of rice of
Thailand and Vietnam are shown in Table 2. The key point to note is that the ADF t-statistic
is -4.79, -4.97 and -3.42 for high quality, medium quality and lower quality rice indicating
that the null of no cointegration can be rejected (given that the critical value is -3.40), which
implies that both the Thai and Vietnam prices across different rice qualities are cointegrated.
The residuals of (1) are then estimated in the form of the TAR and M-TAR models. The
results of the high quality rice are discussed first. Considering the TAR model, the point
estimates are calculated to be γ 1 = −0.42 and γ 2 = −0.25 which have the correct signs for
convergence. The statistic Φ = 12.34 is greater than the 1% critical value implying that the
null hypothesis of no cointegration can be soundly rejected. Given that we find cointegration,
the null hypothesis of symmetric adjustment can be tested using the standard F distribution.
The sample value of F = 1.60 has a p-value of 0.20 implying that we cannot reject the null of
symmetric adjustment. Turning to the M-TAR consistent model, the point estimates are found
to be γ 1 = −0.27 and γ 2 = −0.58 suggesting convergence. The statistic Φ = 13.12 allows us
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
85
to reject the null hypothesis of no cointegration at the 5% significance level. The null
hypothesis of symmetric adjustment provides a sample value of F = 2.87 with a p-value of
0.09 implying that we can reject the null of symmetric adjustment at the 10% significance
level. Moving on to the medium quality rice, we find for the TAR model that γ 1 = −0.29 and
γ 2 = −0.41 suggesting convergence and the null hypothesis of no cointegration reports an
estimate of 12.7 which leads us to reject the null at the 1% significance level. The null
hypothesis of no asymmetry cannot be rejected given the p-value of 0.41. However, the MTAR model reveals a different picture. The point estimates have the correct signs of
convergence, that is γ 1 = −0.30 and γ 2 = −0.52 , and the Φ statistic denoting the null
hypothesis of no cointegration is estimated to be 14.76 which is soundly rejected at the 1%
significance level. However, the null of no asymmetry can be rejected given the p-value of
0.04. Finally, for low quality rice we find that using the TAR and M-TAR framework we can
reject the null hypothesis of no cointegration at the 5% significance level. However, we
cannot reject the null hypothesis of no asymmetry. Finally, the Ljung Box Q statistic shows
that none of the models suffer from problems of serial correlation.
Where asymmetry is found to exist, in this case the high and medium quality market, the
point estimates of γ 1 and γ 2 are found to be negative, suggesting convergence for the MTAR model. Since | γ 1 | < | γ 2 | , the M-TAR model exhibits little adjustment for positive
Δε t −1 but substantial decay for negative Δε t −1 . In other words, increases tend to persist, but
decreases tend to revert quickly back to the attractor. Thus the results suggest that using the
M-TAR consistent model we can find evidence of asymmetric adjustment. The finding of
asymmetric adjustment leads us to conclude that the results generated by the M-TAR model
have more power than the Engle Granger tests (Enders 2001). In the case of low quality rice
we find no evidence of asymmetry and thus in this case the results of the standard linear
Engle-Granger test will have a higher power than the threshold models.
Given the finding of cointegration an ECM is estimated. For the high and medium quality
rice market we estimate an ECM with threshold adjustment and a simple ECM is estimated
for the lower quality rice where we find evidence of symmetric adjustment. The ECM allows
us to nest the short and long run dynamics and to determine, using the concept of weak
exogeniety, whether one country’s exports lead the other country’s exports or whether there
is feedback between the two exporters. The results for the ECM are reported in Table 3
below.
In the high quality segment, the t-ratios of the error correction terms suggest the adjustment
to long run equilibrium is slow but significant at the 5% level for Vietnam rice export prices.
The error correction terms for the Thai export price are not significant at the 5% level. The
weak exogeneity tests given by the p-values in the last row of Table 3 confirm that the Thai
export price evolves independently and the Vietnam export prices adjust to any deviation
from long run equilibrium. One may conclude that the Thai export prices act as a price leader
and Vietnam adjusts its export prices taking the Thai export prices as a reference point. There
is some evidence at the 5% significance level that Vietnam prices Granger cause Thai prices
in the short run. Moving on to the medium quality rice segment, we find the error correction
terms to be significant and that the adjustment is relatively slow when prices are increasing
than when they are decreasing. The weak exogeneity tests suggest that there is no single price
that is driving the other and therefore we can conclude that there is price feedback. In the
short run there is no evidence of Granger causality. Finally, in the low quality rice segment,
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
86
we find the error correction term for the Vietnam price equation to be significant. This
suggests that the Vietnam price adjusts to any deviation form the long run equilibrium and
the Thai export prices evolve independently over time. One can conclude that the Thai prices
act as a price leader driving the Vietnam rice export prices. The weak exogeneity tests
confirm these results. In this segment the Vietnam export prices Granger cause Thai prices
and there is no evidence of Granger causality from Thai to Vietnam export prices. For all the
ECM’s the Ljung Box Q statistic reveals that none of the models suffer from serial
correlation.
4. Conclusion
A method due to Enders and Siklos (2001) was employed to test the hypothesis of
cointegration with asymmetric adjustment, and was applied to the Thai and Vietnam rice
export prices; both countries being leading exporters of rice. The method extends the Engle
Granger procedure allowing for either TAR or momentum TAR adjustment toward the
cointegrating vector. If asymmetry exists, the power of the M-TAR test is higher than that of
the Engle Granger test (Enders 2001).
The results of the Engle Granger procedure indicate that a long run relationship exists
between the rice export prices of Vietnam and Thailand. This implies that more effective
transmission of market information has led to more stability in export pricing, resulting in
Vietnam’s export prices being closely linked to Thai prices. When testing for asymmetry we
find that for 5% and 15% broken rice there is along run relationship and the underlying
process of adjustment is asymmetric. The nature of asymmetry is captured by the M-TAR
model which suggests that the path of adjustment to the long run equilibrium relation is
relatively faster when the price differential is decreasing than opposed the case when it is
increasing. When prices are decreasing, the gap between the prices decreases at a faster rate
as opposed to the case when both prices are increasing. Dawe (2004) states that there has
been a marked shift from low quality to high quality rice because of the rising levels of
income. The results reflect the increased competition from Vietnam in the higher quality rice
market which may have resulted in the price differential narrowing over time (Dawe 2004).
Another possible explanation for this asymmetry is that Vietnam may still be known for
selling low quality rice (as it traditionally has been an exporter of low quality rice) even
though in recent years Vietnam has improved milling facilities. Besides, Vietnam entered the
market when prices were steadily declining (Nielsen 2003) and therefore Vietnam prices in
the high and medium quality rice market would have to decrease at a faster rate in
comparison to Thailand to maintain its market share as a major exporter of high and medium
quality rice. The finding of this asymmetric pattern of adjustment also lends some support to
the finding by Yumkella et. al., (1994) that the rice market is characterized by imperfect
competition in the high quality market. When considering the lower quality market, that is
35% broken rice for Thailand and 25% broken rice for Vietnam, we find that the process of
adjustment is symmetric suggesting that price increases are transmitted to the same extent as
price decreases. This implies that the competition between Vietnam and Thailand is quite
high for the low quality rice market. Vietnam gained prominence as a major exporter of low
quality rice and therefore major importers of low quality rice, such as Indonesia, have
imported almost equal amounts of rice from both these two exporters. The finding of
symmetric price adjusting behavior lends support to the intense competition for markets by
Vietnam and Thailand in the low quality market. Overall, the finding of the different nature
of price dynamics for different qualities of rice, indicate that no uniform policy should be
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
87
applied to the different qualities of rice. Further research may be made to investigate the
underlying causes and policy implications of the different dynamic behavior of prices across
different qualities of rice.
Endnotes
1. Source: World Grain Situation, USDA.
2. Authors calculation; based on data from the Thailand Grain and Feed Weekly Rice Price
Update from the U.S. Embassy in Bangkok and the Creed Rice Market Report, Creed
Rice Co. Inc., Houston, Texas, USA.
3. If ωt is not white noise, an Augmented Dickey Fuller (ADF) test may be used where
lagged values of Δεˆt may be added to (2).
4. To utilize Chan’s methodology, the estimated residual series was sorted in ascending order,
that is, εˆ1 < εˆ2 < ... < εˆT where T denotes the number of usable observations. According to
the method, the largest and smallest 15% of the εˆt series were eliminated and each of the
remaining 70% of the values were considered as possible thresholds. For each of the
possible thresholds the equation was estimated using (3) and (4). The estimated threshold
yielding the lowest residual sum of squares was deemed to be the appropriate estimate of
the threshold.
5. The estimated residual series was sorted in ascending order, that is, Δεˆ1 < Δεˆ2 < ... < ΔεˆT
where T denotes the number of usable observations. As before with the TAR model, the
largest and smallest 15% of the εˆt series were eliminated and each of the remaining 70%
of the values were considered as possible thresholds. For each of the possible thresholds
the equation was estimated using (3) and (5). The estimated threshold yielding the lowest
residual sum of squares was deemed to be the appropriate estimate of the threshold.
References
Chan, K. S. 1993. “Consistency and Limiting Distribution of the Least Squares Estimator of
a Threshold Autoregressive Model,” The Annals of Statistics, 21, 520-533.
Chen, C., B. A. McCarl, and C. Chang. 2006. “Estimating the Impacts of Government
Interventions in the International Rice Market,” Canadian Journal of Agricultural Economics
54, 81-100.
Cramer, G., E. J. Wailes, and S. Shui. 1993. “Impacts of Liberalizing Trade in the World
Rice Market,” American Journal of Agricultural Economics, 75, 219-226
Dawe, D. 2002. “The changing structure of the world rice market,” 1950-2000. Food Policy,
27, 355-370.
Enders, W. 2001. “Improved critical values for the Enders-Granger unit-root test,” Applied
Economics Letters, 8, 257-261.
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
88
Enders, W. and P. L. Siklos. 2001. “Cointegration and Threshold Adjustment,” Journal of
Business and Economic Statistics 19, 166-176.
Engle, R. F. and C. W. J Granger. 1987. “Cointegration and Error Correction:
Representation, Estimation and Testing,” Econometrica 49, 251-276.
Ghoshray, A. 2002. “Asymmetric Price Adjustment and the World Wheat Market,” Journal
of Agricultural Economics, 53, 299-311.
Goletti, F. and S. Babu 1994. “Market Liberalization and Integration of Sorghum Markets in
Malawi,” Agricultural Economics 11, 311-324.
Karp, L. and J. M. Perloff. 1989. “Dynamic Oligipoly in the Rice Export Market,” Review
of Economics and Statistics, Vol. LXXI, 462-470.
Minot, N. and F. Goletti. 1998. “Export Liberalization and Household Welfare: The Case of
Rice in Vietnam” American Journal of Agricultural Economics. 80, 738-749.
Nielsen, C. P. 2003. “Vietnam’s Rice Policy: Recent Reforms and Future Opportunities,”
Asian Economic Journal 17, 1-26.
Petrucelli, J. and S. Woolford 1984. “A Threshold AR(1) Model,” Journal of Applied
Probability 21, 270-286.
Rakotoarisoa, M. A. 2006. “Policy Distortions in the Segmented Rice Market,” International
Food Policy Research Institute, May 2006. MTID Discussion Paper No. 94.
Siamwalla, A. and S. Heykin. 1983. “The World Rice Market: Structure, Conduct and
Performance,” International Food Policy Research Institute, Research Report No. 39.
Tong, H. 1990. Non Linear Time Series: A Dynamical Approach. Oxford, UK: OUP.
Yumkella, K. K., L. J. Unnevehr, and P. Garcia. 1994. “Non-competitive Pricing and
Exchange Rate Pass Through in Selected U.S. and Thai Rice Markets,” Journal of
Agricultural and Applied Economics, 26, 6-16.
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
Table 1: Unit Root Tests
Prices
ADF
Levels
Thailand (15% Broken)
Thailand (5% Broken)
Thailand (35% Broken)
Vietnam (25% Broken)
Vietnam (15% Broken)
Vietnam (5% Broken)
89
-1.21 (1)
-1.23 (1)
-1.04 (1)
-1.00 (1)
-1.20 (1)
-1.25 (1)
Differences
-7.73** (0)
-7.89** (0)
-7.44** (0)
-7.34** (0)
-7.30**(0)
-7.60**(0)
ERS
Levels
-1.12 (1)
-1.09 (1)
-1.03 (1)
-1.09 (1)
-1.24 (1)
-1.23 (1)
Ng-Perron
Levels
-1.13 (1)
-1.10 (1)
-1.04 (1)
-1.21 (1)
-1.33 (1)
-1.29 (1)
** and * Denote significance at the 1% and 5% level respectively. The numbers in parentheses denote lag lengths.
Table 2: Cointegration Tests for Thai and Vietnam Rice Prices
Thai-Vietnam (5% Broken)
Thai-Vietnam (15% Broken)
Engle
Consistent Consistent Engle
Consistent Consistent
Granger
TAR
M-TAR
Granger
TAR
M-TAR
-0.34
-0.42
-0.27
-0.36
-0.29
-0.30
γ1
(4.79)
(4.49)
(3.86)
(4.97)
(2.88)
(3.87)
N/A
-0.25
-0.58
N/A
-0.41
-0.52
γ2
(2.46)
(3.72)
(4.34)
(4.15)
N/A
12.34**
13.12**
N/A
12.7**
14.76**
Φ
N/A
1.60 [0.20] 2.87 [0.09] N/A
0.69 [0.41] 4.05 [0.04]
γ1 = γ 2
Thai (35%)-Vietnam (25%)
Engle
Consistent Consistent
Granger
TAR
M-TAR
-0.21
-0.16
-0.11
(3.42)
(1.85)
(1.11)
N/A
-0.26
-0.28
(2.96)
(3.51)
N/A
6.12*
6.78*
N/A
0.55 [0.45] 1.76 [0.18]
Q
6.24 [0.18]
1.06 [0.89]
0.73 [0.94]
1.14 [0.83]
0.37 [0.98]
0.44 [0.97]
0.41 [0.98]
6.66 [0.15]
7.14 [0.12]
Note: the values corresponding to Φ are compared with the Φ tables computed by Enders and Siklos (2001). ** and * denote the 1% and 5% significance level
respectively. The numbers in parentheses denote t-values. For Null hypothesis of symmetry and the Q statistics in the last two rows, the numbers in the square brackets
denote p-values.
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
90
Table 3: Error Correction Model
Δ Viet5(t-1)
Δ Thai5(t-1)
Δ Viet15(t-1)
Δ Thai15(t-1)
Δ Viet25(t-1)
Δ Thai35(t-1)
ECT (+)
ECT (–)
ECT
Q
H 0 (weakly exog)
Δ Thai5(t)
0.19 (1.95)
0.18 (1.89)
High Quality
Δ Viet5(t)
0.47 (4.17)
-0.10 (-0.96)
Medium Quality
Δ Thai15(t)
Δ Viet15(t)
0.14 (1.40)
0.22 (2.27)
Low Quality
Δ Viet25(t)
0.50 (4.37)
-0.09 (-0.80)
0.13 (1.86)
0.20 (1.39)
-0.18 (-2.23)
-0.40 (-2.49)
0.12 (1.77)
0.32 (2.21)
-0.18 (2.28)
-0.36 (2.24)
5.67 [0.22]
2.45 [0.09]
3.97 [0.41]
5.04 [0.01]
6.89 [0.14]
3.60 [0.03]
1.77 [0.77]
4.59 [0.01]
The numbers in parentheses denote t-values and the numbers in square brackets denote p-values.
Δ Thai35(t)
0.24 (2.56)
0.14 (1.34)
0.43 (3.59)
-0.08 (-0.64)
0.05 (0.87)
5.34 [0.25]
0.77 [0.38]
-0.21 (2.72)
2.83 [0.58]
7.43 [0.00]
Ghoshray, International Journal of Applied Economics, September 2008, 5(2), 80-91
Figure 1
91